Embodiments of the present invention generally relate to decision support tools, and more particularly relate to a faceted browser for decision-making that provides interactive visualization of a decision within a complex network of sub-decisions (i.e. a “decision space”) as choices are made.
Everyday, modern culture asks us to make myriad decisions. How do we decide which college to attend, which job to take, or which restaurant to eat at? We typically rank the alternatives along relevant dimensions, or facets. Each facet, in turn, includes a plurality of choices. When choosing a restaurant, for example, facets may include type of cuisine, average meal price, and distance from our home. The choices for type of cuisine may include Italian, French, Chinese, or Korean. We use our understanding of the facets to weigh high-ranking alternatives. For example, the closest restaurant may be the most expensive, but nevertheless may be the best choice if it serves the cuisine we want.
When decisions become complex, decision support tools can facilitate the decision-making process. A decision support tool is a computer-implemented system that models the groups of connected choices (i.e., facets) involved in making complex decisions as a decision space. Decision support tools help users navigate through a decision space to reach an optimal result. Each forward step is a choice that limits the number of remaining alternatives in the result set, while backtracking steps expand the number of alternatives.
A typical decision support tool models a decision space as a single spanning tree or hierarchy. The hierarchy is often mapped to a user interface, allowing users to make decisions by navigating the hierarchy. One example is a YELLOW PAGES telephone directory, which is a two-tier hierarchy of local businesses organized by business type and name.
A single hierarchy limits the usefulness of a decision support tool in several ways. First, users must navigate from the top of the hierarchy down to the bottom. With respect to the telephone directory, a user must select a business type and then a name; navigation in the opposite direction is not supported. Second, sufficient knowledge is needed to make choices. In particular, users must be able to understand the terminology used to label each step. Third, a single hierarchy cannot support users who may be approaching a decision from different contexts. For example, the telephone directory will not be useful to a user who only knows the location of a business.
Faceted browsers are an alternative to single-hierarchy decision support tools. Faceted browsers allow users to navigate multiple hierarchies in any order. For example, if the decision is “What graph type should I use to plot a particular series of data?” one facet may organize graph types by tasks, such as “compare trends” or “show percentages.” A second facet may organize graph types by the structure of the data to be plotted, such as hierarchical or non-hierarchical. Users can start the process of selecting a graph type by choosing a task, data structure type, or any other facet. Because facets can be chosen in any order, there is a higher chance that users with a wide variety of contexts and knowledge will be able to use the decision support tool successfully. If an analyst is familiar with the data to be plotted, she might start with the data structure facet, while a usability engineer might start with the task facet.
Although faceted browsers represent an improvement in usability over single-hierarchy decision support tools, they still have usability problems. Users are still likely to encounter choices between concepts that they do not understand. When users do not understand the logical grouping for a facet or the particular terms used to label a choice, they are unable to anticipate how making that choice will affect the decision space they are traversing. In some cases making a choice may prune too much of the decision space, while in other cases it may prune too little. Additionally, a choice taken by misinterpreting its label may move the user father away from the optimal result rather than closer, perhaps making that result unattainable.
Current approaches to the above problems include using indicators (e.g., mouse-over tool-tips and text labels) to show how a particular choice might affect metrics associated with the result set (e.g., result set size and result set sub-categories). However, merely knowing the degree to which the decision space will be pruned does not convey much information about how the decision space is organized and which parts of the decision space will be pruned as a result of the choice. This information is particularly important in helping users understand the logic behind unfamiliar facets and the meaning of unfamiliar choice labels.